Background
-
Does professional communication improve health outcomes? What professional communication network properties are associated with health outcomes?
-
What methods have been used for which types of research questions?
-
What are the main limitations of the SNA methods?
-
What is the quality of these studies?
-
What is the quantity of SNA studies? What was the evolution over time?
-
To what extent has this research taken place in low- and middle-income countries?
-
To what extent has this research focused on community-based health providers?
Methods
Definitions
Search strategy
Concept 1: social network analysis
Concept 2: diffusion of innovations
Concept 3: knowledge translation and transfer
Study inclusion and exclusion criteria
-
Does the study use SNA methods?
-
Are the study subjects healthcare providers?
-
Is the communication/relationship of interest between healthcare providers?
-
Does the research focus on professional communication?
-
Is there some metric used for performance, defined as assessing patient outcomes?
Study quality assessment
Data extraction strategy
Data synthesis and presentation
Results
Year of publication | Number | Percent |
---|---|---|
1990–2000 | 0 | 0% |
2000–2010 | 0 | 0% |
2010–2016 | 6 | 100% |
Country | ||
Australia | 1 | 17% |
USA | 5 | 83% |
Type of health facility | ||
Hemodialysis center | 1 | 17% |
Hospital-based | 3 | 50% |
Nursing home | 1 | 17% |
Primary health care (PHC) | 1 | 17% |
Type of health professional | ||
Multidisciplinary teams | 4 | 67% |
Nursing staff | 2 | 33% |
Type of patients | ||
Emergency and outpatient department patients | 1 | 17% |
Hemodialysis patients | 1 | 17% |
Medical-surgical unit patients | 1 | 17% |
Nursing home patients | 1 | 17% |
PHC patients with alcoholism | 1 | 17% |
Renal and respiratory ward patients | 1 | 17% |
Study design | ||
Experimental | 1 | 17% |
Observational | 5 | 83% |
Mixed methods | 2 | 33% |
Quantitative only | 4 | 67% |
Qualitative only | 0 | 0% |
Primary research question
What SNA methods have been used to study professional communication and performance among healthcare providers?
Author (date)country | Study objectives | Research questions | Study design | Data collection method(s) | Critical appraisal(0/+/++) |
---|---|---|---|---|---|
Type/number of healthcare worker (HCW) | Patient outcomes | Study findings | Limitations | ||
Effken et al. (2011) [45] USA |
Study objectives: Identifying nursing unit communication patterns associated with patient safety and quality outcomes |
Research questions: Can ORA’s visualizations be used to identify patient care unit network communication patterns that affect patient safety and quality outcomes?Do unit network characteristics differ by shift?What network characteristics measured by ORA metrics are related to specific safety and quality measures? |
Study design: Observational;cross-sectional |
Data collection method(s): Organizational network Analysis questionnaire, previously collected survey with patient outcome data |
Critical appraisal: ++
|
Type/number of HCW: Nursing staff, number not stated |
Patient outcomes: Adverse drug events, Falls, Symptom management difference, Symptom management capacity, Simple self-care management, Complex self-care management |
Study findings: Demonstrated utility of ORA software for healthcare research and relationship of nursing unit communication patterns to patient safety and outcomes. Differences between day and night shift communication networks. Found more communication not always associated with better patient outcomes, specifically falls and adverse drug events |
Limitations: Small and homogenous sample size. Only nursing staff. Limited to weekday shifts. Some of the patient outcomes (falls and adverse drug events) were infrequent events | ||
Lindberg et al. (2013) [47] USA |
Study objectives: Evaluate how intervention affected adherence to infection prevention protocols, patient outcomes, and dialysis center social networks |
Research questions: Does a package of interventions including membership to a collaborative emphasizing positive deviance change HCW collaboration, infection prevention, and innovation networks? Do patient outcomes change? |
Study design: Experimental pre-post intervention Longitudinal Mixed-methods SNA retrospective |
Data collection method(s): Survey, focus group discussions, observation, and patient data extraction |
Critical appraisal: +
|
Type/number of HCW: Multidisciplinary staff at an outpatient hemodialysis facilitySNA 51 identified (46 completed each of the 2 surveys 90%)FGD 16 |
Patient outcomes: Infection rates in patients |
Study findings: There were changes in all three networks following the implementation of the package of interventions.For collaboration: centralization and reach increased, connectivity decreased, and no change in inclusion.For bloodstream infection (BSI): prevention reach increased, the others did not change significantly.For innovation: inclusion and reach increased, reach decreased, and centralization did not change.Qualitative data supports the noted changes in network with staff looking to each other for innovations in infection prevention, working more as a cohesive team.Patient outcomes improved, with lower incidence of BSIs, although they were a relatively rare event. |
Limitations: Results are based on one dialysis center and may not be generalizable to other centers. SNA was based on a retrospective survey and might have been subject to recall bias.The results of the time series analysis are limited as access-related bloodstream infections (AR-BSIs) are a relatively rare outcome and there were a small number of time points between interventions.Researchers were unable to stratify AR-BSIs by access type before 2009. | ||
Alexander et al. (2015) [43] USA |
Study objectives: To evaluate how differences in IT sophistication in nursing homes impact communication and use of technology and associations with skin care and pressure ulcers |
Research questions: What communication strategies do nursing home staff use to provide care to residents at risk of skin breakdown and pressure ulcers? What evidence-based pressure ulcer preventions are used by nursing home staff with diverse IT sophistication? What social networks of CNAs enhance or interrupt workflow and have positive or negative effects on nursing work? |
Study design: Observational mixed-method case studies |
Data collection method(s): For SNA, observation of communication among HCWs was documented using a structured field note guide. |
Critical appraisal: +
|
Type/number of HCW: Nursing staff, FGD, 21;SNA, nurses at 2 nursing homes, 1386 observations (unit of analysis was not based on number of HCWs) |
Patient outcomes: Incidence of pressure ulcers |
Study findings: High IT sophistication lead to more diverse locations for HCW interactions. Low IT sophistication required more face to face interaction in more centralized locations within the nursing home. Patient outcomes captured were more or less equivalent between the two facilities. |
Limitations: The study focused on observations only during the day shift. Individual RNs/LPNs and CNAs were not uniquely identified during observations, and the analysis lumped them together as RNs/LPNs and CNAs, respectively. Two nursing homes with a specific degree of IT sophistication were compared, rather than following any change due to the introduction of IT sophistication. Confounding variables offer an opportunity for increasing bias in the results. Generalizability may not be appropriate as this study was an in-depth analysis of two nursing homes in one state––Missouri. | ||
Creswick and Westbrook (2015) [44] Australia |
Study objectives: Determine if there are network property differences in prescription advice-seeking associated with prescription errors |
Research questions:
1. Identify and measure from whom hospital clinical staff seek medication advice on a weekly basis 2. Quantify the use of other sources of medication information, assess the difference in medication advice-seeking patterns across professional groups 3.Examine network characteristics in relation to prescribing error rates |
Study design: Observational Cross-sectional |
Data collection method(s): Questionnaire for SNA and clinical audit |
Critical appraisal: ++
|
Type/number of HCW: Multidisciplinary: physicians, nurses, and allied health professionals101 participants |
Patient outcomes: Prescription error rates |
Study findings: Limited interprofessional advice-seeking overall (particularly between physicians and nurses). Hubs of advice provisions include pharmacists, junior physicians, and senior nurses. Senior physicians are not involved in these advice exchange networks. The ward with the stronger (denser) advice-seeking network had lower rates of procedural and clinical prescribing errors. |
Limitations: Limited in scope (only two wards).No psychometric assessments. Networks only examined at one point in time. | ||
Mundt et al. (2015) [48] USA |
Study objectives: To understand what team communication structures contribute to alcohol-related utilization of care and medical costs. |
Research questions: What primary care team communication networks are associated with alcohol-related utilization of care and medical costs for primary care patients? |
Study design: Observational Retrospective |
Data collection method(s): Questionnaire administered in person, electronic health record extractions |
Critical appraisal: ++
|
Type/number of HCW: Multidisciplinary: physicians, physician assistants, nurse practitioners, registered nurses, medical assistants, licensed practical nurses, laboratory technicians, radiology technicians, clinic managers, medical receptionists, and other patient care staff.One hundred sixty HCWs were invited, 155 took part; 31 care teams |
Patient outcomes: Alcohol-related emergency department visits,Hospital days and associated costs |
Study findings: Teams’ variations in communication patterns (face to face and through electronic health record) are associated with statistically significant differences in alcohol-related patient utilization and medical costs in their patient panels. Excessive alcohol-using patients may fair better if they are cared for by teams with RNs who interact with more team members including LPNs/MAs and by teams whose frequent daily face-to-face communication to the primary care practitioner has been streamlined to a smaller number of team members. |
Limitations: Only six practices in limited geography included. No information on content of communication. No information on frequency and quality of alcohol services delivered. Unclear rationale for type of communication method. Increased risk of type I error. Study may underestimate full impact given underreporting of alcohol-related diagnoses in electronic health record. | ||
Hossain and Guan (2012) [46] United States |
Study objectives: To understand coordination in an emergency department through measures ofperformance and quality |
Research questions: Test the following hypotheses: Performance of coordination in the emergency department is influenced by the social network. Performance of coordination in the emergency department is influenced by the centrality of the network. Performance of coordination in the emergency department is influenced by the density of the network. Performance of coordination in the emergency department is influences by the degree of connections in the network. |
Study design: Observational Cross-sectional |
Data collection method(s): National Hospital Ambulatory Medical Care Survey (NHAMCS), patient record surveys selected from emergency departments |
Critical appraisal: ++
|
Type/number of HCW: Multidisciplinary: emergency department hospital staff. Staff included in patient reports from 359 emergency departments |
Patient outcomes: Length of visit, wait time to see physician, revisits within 72 h, deaths within emergency department, and left before seeing physician |
Study findings: Coordination and the social network are heavily related within the emergency department. Specifically, as emergency department network density increases, number of patients waiting over triage time decreases but does not influence average wait times. As degree of connection increases, the wait time for patients increases. No evidence of connection between quality of service and death and the social networks. Quality of coordination in emergency department is influenced by centrality of the network. As communication in emergency department increases, the number of patients revisiting decreases. |
Limitations: NHAMCS dataset is incomplete, contains less than 40 surveys of each emergency department, which is less than the assumed volume of patients in a 3-month period. |
Authors | Effken et al. | Hossain and Guan | Lindberg et al. | Alexander et al. | Creswick and Westbrook | Mundt et al. |
---|---|---|---|---|---|---|
Data collection method | Network survey, previously collected survey with patient outcomes | Extraction from National Hospital Ambulatory Medical Care Survey (NHAMCS), patient record surveys selected from emergency departments | Survey, focus group discussions, observation, and patient data extraction | Observation, previously collected survey | Network survey and clinical audit | Network survey and electronic health record extractions |
Boundary specification method/sampling (if applicable) | All nursing staff who worked in one of seven patient care units in three magnet hospitals | Emergency departments of 359 hospitals responded to the ambulatory survey section of NHAMCS survey conducted by the CDC. | All staff at 21 hemodialysis facilities that form part of the CDC Hemodialysis BSI Prevention Collaborative | Comparative case study of two units within two nursing homes, one with the highest IT sophistication and one with the lowest IT sophistication based on a statewide census in 2007. Nodes were both HCWs and the locations and content of their interactions. | All HCWs in two wards | Eight clinics in Southern Wisconsin were invited to participatein the study, and six agreed. Sites were chosen based on consultation with leadership from the healthcare system. |
Network category studied 1. Whole, ego, or hybrid network 2. Directed or undirected 3. Valued or dichotomous | 1. Whole network 2. Directed 3. Valued | 1. Whole network 2. Directed 3. Dichotomous | 1. Whole network 2. Directed 3. Valued | 1. Whole network 2. Directed 3. Valued | 1. Whole network 2. Directed 3. Valued | 1. Whole network 2. Directed 3. Valued |
Response rate | Not stated | N/A as SNA data extracted from surveys on patients | 90% | N/A as SNA from observation | 90% | 97% |
Network metrics used | Clustering coefficient, component count strong, component count weak, density, diffusion, fragmentation, hierarchy, isolates, in-degree centrality, out-degree centrality, eigenvector centrality, simmelian ties, betweenness centrality, number of triads, and number of cliques | SNA metrics: degree, density, and centrality | Connectivity, inclusion, reach, and centralization | None | Density Reciprocity In-degree centrality | In-degree centrality, tie strength |
Analyses conducted | Correlations (Spearman Rho) calculated between SNA metrics and patient outcomes | Multiple linear regression, p values and r values reported | Quantitative: Pearson X
2 and Fisher’s exact test, t test. Reported p valuesQualitative analysis: reflexive observation and contextual analysis | Quantitative: calculated highest and lowest ITS NH from survey data in an earlier study Qualitative: axial coding, themes developed using human factors theory | Chi-squared with p values | Linear modeling (GLMM) and sensitivity analyses |
Software | ORA, Excel | UCINET, SPSS, Excel | Not stated | ORA, Nvivo, Excel | UCINET and NetDraw | UCINET, HLM 7.0 |
Network map (yes/no) | Yes | Yes | Not stated | Yes | Yes | Yes |
Further research | Replicate study, expand to larger, more diverse group of patient care units. Consider shifting to more patient-centric focus, including full team of care providers | Further research needed to verify the relationship suggested by this study between coordination and social network analysis. Survey of emergency departments within Australia for a period of 1 year, to allow accurate measurements to be taken and utilized for the study and for verifying the relationship between social networks and coordination in an emergency department. | None stated | To demonstrate how organization analytics about communication can be used to benchmark evidence-based practices | Further research on link between medication advice-seeking networks and errors, as this study suggests. Also, whether the increased use of electronic medication management systems means that information needs are met through channels other than communication between physicians, nurses and pharmacists, or that information sharing regarding medication issues is reduced and may impact medication safety. Evaluate interventions to engage senior physicians in advice exchange networks. Further health applications of SNA surveys needed to improve validity and reliability of tools. | Longitudinal and experimental studies needed to explore the causal pathways between team communication variables and alcohol-related patient care |
Network intervention (yes/no) | No | No | Yes (although intervention not based on baseline network analysis. Rather, it was developed with the intention of changing HCW networks) | No | No | No |
Social Network Analysis Metric | Effken et al. | Lindberg et al. | Alexander et al. | Creswick and Westbrook | Mundt et al. | Hossain and Guan | Total |
---|---|---|---|---|---|---|---|
Centralization | X | 1 | |||||
Centrality (in-degree) | X | X | X | 3 | |||
Centrality (out-degree) | X | 1 | |||||
Centrality (eigenvector) | X | 1 | |||||
Centrality (betweenness) | X | 1 | |||||
Clustering coefficient | X | 1 | |||||
Component count strong | X | 1 | |||||
Component count weak | X | 1 | |||||
Connectivity | X | 1 | |||||
Degree | X | 1 | |||||
Density | X | X | X | 3 | |||
Diffusion | X | 1 | |||||
Fragmentation | X | 1 | |||||
Hierarchy | X | 1 | |||||
Inclusion | X | 1 | |||||
Isolates | X | 1 | |||||
Number of triads | X | 1 | |||||
Number of cliques | X | 1 | |||||
Reach | X | 1 | |||||
Reciprocity | X | 1 | |||||
Simmelian ties | X | 1 | |||||
Tie strength | X | 1 | |||||
Total | 15 | 4 | 0 | 2 | 2 | 3 | 26 |
Metric
|
Study
|
Patient outcomes
|
Association with metric
|
Overall association
|
Centrality | Effken et al. | Adverse drug events (ADEs) | “Betweenness centrality” positively correlated (rho = .73) with ADEs | Generally, as centrality measures increase, patient outcomes improve; however, there were many patient outcomes for which there was no significant association with a centrality measure. Effken exception. Higher betweenness centrality, with potentially more gatekeepers resulted in more ADEs. With symptom management difference, the seemingly inconsistent association with centrality could actually point to the importance of small group communication with this outcome measure and that those with more out-degree ties are novices seeking advice. |
Falls | Not significant | |||
Symptom management difference | “Centrality out-degree” negatively correlated (rho = −.79) although eigenvector centrality positively correlated (rho = .69) | |||
Symptom management capacity | Not significant | |||
Simple self-care management | Not significant | |||
Complex self-care management | Not significant | |||
Lindberg et al. | Access-related bloodstream infections | Not significant | ||
Mundt et al. | Alcohol-related emergency department visits | Statistically significant (sig.) GLMM model with only weak “in-degree ties” had positive association(RR 1.23, p < 0.01), models with any strong ties had inverse association (RR range 0.8–0.9, p < 0.05) | ||
Alcohol-related hospitalizations | Sig. GLMM models with groups of HCWs with any weak “in-degree ties” had positive association (RR 1.1, p < 0.05, RR 1.25, p < 0.01), model with groups of HCWs with only strong ties had inverse association (RR .95, p < 0.05) | |||
Alcohol-related costs per 1000 team patients over 12 months | In an average team size of 19, the addition of a HCW with strong “in-degree ties” reduced cost by $1030 (p < 0.05),weak ties increased cost by $2922 (p < 0.01) | |||
Hossain and Guan | Wait time to see physician | Not significant | ||
Revisits within 72 h | Not significant | |||
Deaths within emergency department | Not significant | |||
Left before seeing physician | “Network centralization” inversely associated (beta = − 0.221, sig. < 0.001) | |||
Metric
|
Study
|
Patient Outcomes
|
Association with metric
|
Overall association
|
Density | Effken et al. | Adverse drug events | Not significant | Density positively associated with improved patient outcomes. However, there were patient outcomes for which there was no significant relationship with density. |
Falls | Not significant | |||
Symptom management difference | Positively associated (rho = 0.70, p < 0.10) | |||
Symptom management capacity | positively associated (rho = 0.75, p < 0.10) | |||
Simple self-care management | Not significant | |||
Complex self-care management | Not significant | |||
Creswick and Westbrook | Prescription error rates (procedural and clinical) | Inversely associated (ward A error rates 5.46 and 1.81 with density 12% vs ward B error rates 1.53 and 0.63 with density 7%) | ||
Hossain and Guan | Wait time to see physician | Inversely associated (beta = − 0.107) for waiting “overestimated triage time” but not significant for “waiting above average” | ||
Revisits within 72 h | Inversely associated (beta = − 0.159, sig. = 0.003) | |||
Deaths within emergency department | Not significant | |||
Left before seeing physician | Inversely associated (beta = − 0.273, sig. < 0.001) |
Study | Objectives/research questions | Research question categories | Methods |
---|---|---|---|
Effken et al. | Identify nursing unit communication patterns associated with patient safety and quality outcomes1. Can ORA’s visualizations be used to identify patient care unit network communication patterns that affect patient safety and quality outcomes?2. Do unit network characteristics differ by shift?3. What network characteristics measured by ORA metrics are related to specific safety and quality measures? | 1. Descriptive 2. Descriptive 3. Relational |
Design: observational, cross-sectional Data collection: Organizational Network Analysis questionnaire, patient outcome survey Analyses: correlations(Spearman Rho) calculated between SNA metrics and patient outcomes |
Lindberg et al. | Determine if intervention changed adherence to infection prevention protocols, patient outcomes and dialysis center social networks 1. Does a package of interventions including membership to a collaborative emphasizing positive deviance change HCW collaboration, infection prevention and innovation networks? 2. Do patient outcomes change? | 1. Causal 2. Causal |
Design: experimental, longitudinal, mixed methods, pre-post intervention Data collection: survey, FGD, observation, patient data extraction Analyses: quantitative: Pearson X2 and Fisher’s exact test, t test Reported p valuesQualitative analysis: reflexive observation and contextual analysis |
Alexander et al. | Evaluate how differences in IT sophistication in nursing homes impact communication and use of technology related to skin care and pressure ulcers. 1. What communication strategies do nursing home staff use to provide care to residents at risk of skin breakdown and pressure ulcers? 2. What evidence-based pressure ulcer preventions are used by nursing home staff with diverse IT sophistication? 3. What social networks of CNAs enhance or interrupt workflow and have positive or negative effects on nursing work? | 1. Descriptive 2. Descriptive 3. Relational |
Design: observational mixed methods, case studies Data collection: observation, previously collected survey Analysis: quantitative: calculated highest and lowest ITS NH and patient outcomes from survey data in an earlier study Qualitative: axial coding, themes developed using human factors theory |
Creswick and Westbrook | Determine if there are network property differences in prescription advice-seeking associated with prescription errors 1. Identify and measure from whom hospital clinical staff seek medication advice on a weekly basis 2. Quantify the use of other sources of medication information, assess the difference in medication advice-seeking patterns across professional groups 3. Examine network characteristics in relation to prescribing error rates | 1. Descriptive 2. Descriptive 3. Relational |
Design: observational, cross-sectional Data collection: network survey and clinical audit Analyses: chi-squared with p values |
Mundt et al. | To understand what team communication structures contribute to alcohol-related utilization of care and medical costs 1. What primary care team communication networks are associated with alcohol-related utilization of care and medical costs for primary care patients? | 1. Relational |
Design: observational, cross-sectional, retrospective Data collection: network survey, electronic health record extractions Analyses: linear modeling (GLMM) and sensitivity analyses |
Hossain and Guan | To understand coordination in an emergency department through measures of performance and quality 1. Is performance of coordination in the ED influenced by the social network? 2. Is performance of coordination in the ED influenced by the centrality of the network? 3. Is performance of coordination in the ED influenced by the density of the network? 4. Is performance of coordination in the ED influenced by the degree of connections in the network? | 1. Causal 2. Causal 3. Causal 4. Causal |
Design: observational, cross-sectional Data collection: survey extraction Analyses: multiple linear regression, p values and r values reported |